Thursday, December 04, 2014

Origins

Research Question





Do terrorists target urban environments more now than in past days?



Purely descriptive analysis of a change of time in terrorists targeting behavior

Data Collection

Main Database

  • Open Source Database run by University of Maryland
  • contains more than 120,000 observations on over 120 variables between 1970 and 2013
  • detailed description of target and effect of attack
    • target & weapon,
    • human & property damage
    • perpetrator



The GTD is a powerful database!

plot of chunk unnamed-chunk-2 Figure 1: attacks per year in all world regions

plot of chunk unnamed-chunk-3 Figure 2: attacks per year in all world regions, indicating transition of collectors: PGIS - 1997, CETIS - 2008, ISVG - 2011, START since 2011

Main Database

  • improvable data quality
  • a lot of information is missing



    The GTD is not providing enough information to answer our research question

Main Database

  • Target's potential urbanity is provided
  • Environment's is not
    • Location (suburb? center?)
    • Importance (capital? stronghold?)
    • Urbanity

plot of chunk unnamed-chunk-4

Addtional Databases

Limitations

  • Validity
    • population an insufficient proxy for urbanity
    • but economic, political etc. data not available open source on city level
    • and population probably the single most important one
  • Reliability
    • multiple collection sources over decades with unreproducible methods in unclear administrative environments

Data Cleaning

Challenges

Within single and across datasets

Issue Example Detail Solution

Inconsistency

Spellings

  • "Congo, Kinshasa"
  • "Congo, Dem. Rep."
  • RegEx
  • Hand Coding

Quality

Typos

Errors

  • "Agfhanistan" vs.

"Arabistan"

  • RegEx
  • Hand Coding

Bias

Underreportings

Definitions

  • Soviet Union: 67 vs. Russia: >1200
  • Iraq since 2003 vs. Afghanistan 79-89
  • Control Variables

Hand Coding

  • Adjusting City and Country Names

    X<-gsub("^Teberan","Tehran", X, ignore.case=TRUE)
    X<-gsub("^Teheran","Tehran", X, ignore.case=TRUE)
    X<-gsub("^Teherau","Tehran", X, ignore.case=TRUE)
    X<-gsub("^Tehram","Tehran", X, ignore.case=TRUE)
  • "Reassigning" Cities to Countries

    GTD$country_txt[GTD$country_txt == "Soviet Union" &
                    GTD$city =="Osh" ] <- "Kyrgyzstan"
    GTD$country_txt[GTD$country_txt == "Soviet Union" &
                    GTD$city =="Riga" ] <- "Latvia"
    GTD$country_txt[GTD$country_txt == "Yugoslavia" &
                    GTD$city =="Bogicevica" ] <- "Albania"
    GTD$country_txt[GTD$country_txt == "Yugoslavia" &
                    GTD$city =="Banjica" ] <- "Serbia"

Data Manipulating

Filling, Adding, Extrapolating

  • Geo Referencing absolute & relative city position
  • Proximity Referencing each city by assigning it to its closest Urban Center
  • Establishing Relative City Size as a share of each countries' largest cityies'total population
  • Estimating Past Populations using country level data and growth rates
  • Determining Cities' Rank Size given by the population on a country, regional, and world level
  • Including additional Control Variables for capitals, largest cities, inter-state war, intra-state war, extra-in war, extra-out war, and coastal megacities

Merging to Current Dataset

We merge city level, country level, and GTD data together to a reproducible database containing 114086 observations on 44 variables.

  • Merged database contains NA share of 11%
  • Considered acceptable
  • Potential for further reduction
  • We assume systematic sources

plot of chunk unnamed-chunk-5

plot of chunk unnamed-chunk-6

Data Analysis

Approach

We use a simple linear regression, depending on Relative City Sizes attacked over time for various world regions. We control for

  • Extra-In State War
  • Extra-Out State War
  • Intra State War
  • Inter State War
  • Capitals
  • Target's Potential Urbanity
  • including weightings for human and property damage

Results

Regresison Table

  • There is no general global trend for terrorists moving to urban targets
  • Results vary across world regions:
    • there is a trend in the Arab World
    • there is a trend in Southeast Asia
  • Civil War changes everything!

From Here…

# City Country Attacks (Peace)

1

Santiago

Chile

1426

2

Belfast

United Kingdom

1187

3

Karachi

Pakistan

1155

4

Bogota

Colombia

845

5

Medellin

Colombia

808

# Region Attacks

1

S. Asia

 25854

2

MENA

 25250

3

S. America

 16917

4

W. Europe

 13217

5

C. America

 10158

Human Damage City Country Killed+Injured

1

Baghdad

 Iraq
  4345

2

Lima
 Peru
  2163

3

Karachi

Pakistan

  1879

4

Belfast

United Kingdom

  1706

5

San Salvador

El Salvador

  1491

# Group Country Total Attacks

1

Shining Path

 Peru
  4533

2

FMNLF

El Salvador

  3351

3

Taliban

Afghanistan

  3204

4

IRA

United Kingdom

  2673

5

FARC

Colombia

  2228

Outlook

  • Still a lot of potential for further cleaning
  • More possibilities for including control variables
  • Currently only few projects using GTD for analysis
  • Our approach seems promising for more research endavours
  • First, the Master's Thesis!

References

Cutter, Susan, Douglas B Richardson, and Thomas J Wilbanks. 2003. The Geographical Dimensions of Terrorism. Routledge.

Kilcullen, David. 2013. Out of the Mountains: the coming age of the urban guerrilla. Oxford University Press.

START. 2013. Global Terrorism Database \(GTD\). Maryland: National Consortium for the Study of Terrorism; Responses to Terrorism. http://www.start.umd.edu/gtd.